DE102004029010A1 - Device and method for scattered radiation correction in projection radiography, in particular mammography - Google Patents

Abstract

A device (1) for projection radiography, which is set up for a scattered radiation correction, has an evaluation unit (12) which evaluates scattered radiation distributions stored in a data memory for the scattered radiation correction, which were determined in advance by means of a Monte Carlo simulation. takes into account the multiple interactions of the photons with the object (6) to be examined.

Description

contraption and methods for the scattered radiation correction in the projection radiography, in particular Mammography The invention relates to a device for projection radiography with a radiation source, a detector and a detector downstream evaluation unit, based on the supplied by the detector Projection data, the scattering material distribution of the examined Object approximately determined and dependent from the scatter material distribution from a data storage scatter information and read out the projection data based on the scatter information corrected with regard to the scattered radiation component.

The The invention further relates to a method with scattered radiation correction for the Projection radiography and a method for obtaining scatter information.

Such a device and such methods are known from US 6,104,777 A known.

The in the subject (mamma) generated scattered radiation whose intensity in the Mammography almost the order of magnitude can reach the imaging unscattered direct primary radiation, leads to a deterioration of the image quality, by reducing the contrast, by raising of the noise, and finally in terms of quality of image postprocessing, with which a differentiation various types of tissue, in particular the glandular and adipose tissue in the Mom in imaging is achieved. For differentiation according to Two types of tissues are in mammography techniques with a single Energy spectrum, so with a single voltage of the X-ray tube, or the dual-energy method with two voltage values known. In both make Compensation of scattered radiation is required; in the dual-energy method also because the proportion of scattered radiation in both energy spectra is different.

to Reduction of scattered radiation are already mechanical measures been proposed. The use of slot collimators requires the mechanical displacement of the slot collimators via the Mamma measuring field and is therefore time consuming. Anti-scatter grid does not reduce only the scattered radiation, but also the imaging primary radiation. About arguments, the for the omission of anti-scatter screens speak, there is one since Years of ongoing discussion. At compression thicknesses below 4-5 cm could even reduces the dose or increases the SNR (= signal to noise ratio) if you remove the grid. On the other hand, there are applications where the use of a grid is not technically possible, for example in tomosynthesis.

At computational correction method has already been proposed a variety. From M. DARBOUX, JM DINTEN: Physical model based scatter correction in mammography. In: Proc. SPIE, Vol. 3032, 1997, pages 405 to 410 and from JM DINTEN and JM FULL: Physical model based restoration of mammographies. In. Proc. SPIE, Vol. 3336, 1998, pages 641-650 and in the US 6,104,777 A For example, such methods are known. These are convolution / deconvolution methods, in which a scattered radiation intensity distribution is approximated as a convolution of the primary intensity distribution with suitable convolution kernels. Thus, in the cited documents, an analytical model is proposed with which the physical process of scattering in the scattering object (Mamma) is explicitly calculated as an integral transformation. However, this explicit analytic presentation only describes first-order scattering, not multiple scattering. The intensity distribution of multiply scattered photons is assumed to be a spatially constant background over the detector surface and must be estimated from predefined tables. The analytical model for calculating the first-order scatter contribution requires 4-dimensional numerical integrations (3 spatial coordinates + energy spectrum) for each detector pixel, so it is computationally expensive. Proximions are therefore required to reduce the computational burden. Due to the high computational effort it is proposed to carry out the calculations in advance and to tabulate the results.

Further it is from W. CALENDAR: Monte Carlo calculations of x-ray scatter data for diagnostic radiology. In: Phys. Med. Biol., 1981, Vol. 26, no. 5, pages 835 to 849, Monte Carlo methods for simulation of radiation propagation in radiography.

outgoing From this prior art, the invention is therefore the task to provide a device and method with which a Compared to the prior art, improved stray radiation correction carried out can be.

These Task is performed by a device and the procedures with the in the independent one claims solved characteristics. In dependent on it claims Advantageous embodiments and developments are given.

at the device and the method are in an evaluation analyzed the projection images provided by a detector. First The scattered material distribution is typically attempted in mammography the proportions of glandular and adipose tissue, of an object to be examined approximately determine. In a further processing step, depending on from the scattering material distribution scattering information read out a data store. With the help of the scatter information can then be a correction of the projection images in terms of be made in the projection images scattered radiation component. It is essential that the scatter information read from the data memory in advance by a Monte Carlo simulation have been detected, the multiple interaction of the photons with considered the object to be examined.

The basis for the solution described here is that possible correct physical modeling. In contrast to the state of Technology is a modeling possible the a substantial larger number taken into account by details in the following respect: the occurrence of multiple scattering and Polychromasie and geometrical conditions, in particular peculiarities the scattering distribution at the edges of the object be reproduced. While the scatter correction itself is just a table access, possibly with subsequent interpolation required and the calculation of the scattered radiation distribution in the detector plane reduces to 2-dimensional integrations across the detector plane. In spite of the relatively simple one execution the stray radiation correction, the procedure is not limited to special cases and does not set any drastic simplifications or approximations ahead, such as a simplified acquisition geometry, Monochromaticity of radiation, simplifications of the physical Model or a Taylor evolution according to approximation orders or similar.

at a preferred embodiment the scattering material distribution becomes specific for different ones Image areas of the projection image determined. To carry out the Stray radiation correction in one image area will then be the scattering contributions of the surrounding contributions which depend on the specific distribution of scattered material and corrected accordingly. In this way it is possible to use local Consider variations in scattered radiation.

at a further preferred embodiment become in the area of object borders of the object to be examined scattering information for scattered radiation correction which uses the special geometrical conditions in the area of the object edge consider.

The Scattering information is preferably obtained assuming that the scattering material distribution along the radiation direction is homogeneous. Especially in the context of mammography leads such Assuming only minor deviations from the actual Scatter distribution.

The Calculation of the specific scatter information associated with an image area may continue to be made on the assumption that the Object is also homogeneously structured in the transverse direction to the beam. This simplifies the calculation of the scatter information.

If However, a special high accuracy in the calculation of the scatter information it is asked for, may also be an inhomogeneity taken into account transversely to the beam direction become.

at a preferred embodiment the scattering material distribution is determined by the ratio of incident radiation intensity to the unscattered primary radiation is evaluated in an image area, the values for the primary radiation be determined by a scattering correction based on scatter information which have a characteristic homogeneous scattering material distribution assigned.

The processing steps carried out by the evaluation unit can also be carried out iteratively become. The calculated primary radiation components serve to refine the approximate calculation of the scattered radiation components and in this way to achieve improved values for the primary radiation.

The Stray radiation correction usually does not need to be full resolution be made of the detector. Occasionally it may be enough be to make the scattered radiation correction at selected nodes and between the determined stray radiation correction values at the selected nodes to interpolate.

Further Advantages and embodiments of the invention will become apparent from the following Description forth, in the embodiments The invention will be explained in detail with reference to the accompanying drawings. Show it:

1 the construction of a mammography device, in which a mamma compressed between two compression plates and is illuminated by X-rays;

2 a representation of an assumed for the calculation of the scattered radiation correction, simplified structure of the breast to be examined;

3 a flowchart of a method performed for the scattered radiation correction method;

4 a representation of the assumed tissue distribution of a mamma for the calculation of a simple stray radiation propagation function; and

5 a representation of the assumed for the calculation of a precise scattering radiation function structure of the breast to be examined.

1 shows the construction of a mammography device 1 in which using a radiation source 2 X-rays 3 is produced. The divergence of X-rays 3 is optionally using a collimator 4 limited in 1 is indicated by a single beam aperture. The colli mator 4 However, it may also be such that a multiplicity of x-rays which run almost parallel to one another are produced. Such a collimator 4 may be formed, for example, as a pinhole.

The mammography device 1 also has compression plates 5 between which a mom 6 is compressed. The x-ray radiation 3 passes through the compression plates 5 and the mom 6 and generally traverses an air gap 7 before the X-rays 3 on an x-ray detector 8th that hits a variety of individual detector elements 9 , which includes so-called detector pixels.

The one without interaction with the mom 6 by the mom 6 passing through portion of the X-radiation 3 is also called primary radiation 10 designated. The proportions of X-radiation 3 that after at least one dispersion within the mom 6 on the x-ray detector 8th Secondary radiation, on the other hand 11 called.

It should be noted that under the term scattering any type of interaction between the X-ray 3 and the matter of the mom 6 to be understood, by a change in the direction of propagation of the photons of the X-ray 3 is effected.

As stated at the beginning, the secondary radiation 11 that of the primary radiation 10 pictured structure of the mom 6 can significantly falsify, it is beneficial if the secondary radiation 11 from the X-ray detector 8th taken mega projection pictures 6 can be removed. For this purpose, an X-ray detector leads 8th downstream evaluation unit 12 a stray radiation correction. In order to be able to carry out the scatter correction, model assumptions are made about the structure of the mamma 6 met in 2 are shown. In particular, it is assumed that the tissue structure of the mamma 6 consisting essentially of glandular and adipose tissue, along the propagation direction of the X-ray radiation 3 homogeneous tissue distribution can be described. Accordingly, in 2 into the mom 6 different mammal areas 13 . 14 and 15 drawn, whose differently executed shading different proportions of fat and gland tissue along the propagation direction of the X-ray 3 should illustrate. In the framework of projection radiography this is a simplification that does not lead to serious deviations from the actual litter distribution.

On the basis of this model assumption, a stray radiation correction can now be carried out, the course of which in 3 is shown.

After taking a picture 16 is a projection image 17 that's on the x-ray detector 8th incoming primary radiation 10 and secondary radiation 11 reproduces. The projection image 17 is a data reduction 18 subjected, in the various mammal areas 13 . 14 and 15 each specific tissue distributions are assigned. In addition, information about the geometric relationships, in particular the edges of the mamma 6 be won. With the help of in the data reduction 18 gained information about the physical condition of the mamma 6 can then be in a mamma SBSF atlas 19 one the respective Mammabereich 13 . 14 and 15 assignable scattered radiation propagation function 20 (= Scatter Beam Spread Function = SBSF). With the help of SBSFs 20 and an estimate for the primary radiation 10 can then be a stray radiation correction 21 be performed. The as part of the scattered radiation correction 21 generated correction values can directly on the projection images 17 be applied when the scattered radiation correction for each of the detector pixels 9 of the X-ray detector 8th has been calculated. Due to the small variation of the scattered radiation via the X-ray detector 8th It may be sufficient to make the scattered radiation correction for selected detector areas. These can be individual nodes or groups of detector pixels 9 , The scatter correction for those detector pixels 9 for which no stray radiation correction has yet been determined can then be determined by interpolation 22 be determined, which is a correction image 23 produces the same resolution as the projection image 17 having. By combination 24 of the projection image 17 and the correction image 23 finally results in a finished structure picture 25 , preferably exclusively that of the primary radiation 10 pictured structure of the mom 6 contains.

in the The following is the prerequisite for the radiation correction described here and the ones to be done Processing steps are described in detail:

Requirements:

First, it is assumed that the imaging-critical sensitivity spectrum N (E) is known:
the radiation of the X-ray tubes is polychromatic, wherein the energy spectrum Q _u (E) of the photons emitted as Bremsstrahlung at the anode depends on the applied high voltage U, with which the electrons are accelerated from the cathode to the anode; the maximum photon energy is then E _max (U) = U (keV / kV) = eU; However, not only the emission spectrum is decisive for the imaging, but also the transparency of the spectral filters W (E) used and the spectral sensitivity η _D (E) of the detector 8th , The resulting (normalized) spectral distribution is defined by: N U (E) = Q U (E) W (E) η D (E) / c U , (#1)

wird

With the standardization factor

becomes

mit

H

Schichtdicke der Mamma 6

x_G

Schichtdicke Drüsengewebe/cm

x_F = H – x_G

Schichtdicke Fettgewebe/cm

ρ_G, ρ_F

Dichte Drüsen- bzw. Fettgewebe [g/cm³]

b_G = ρ_Gx_G

Massenbelegung Drüsengewebe [g/cm²]

b_F = ρ_Fx_F

Massenbelegung Fettgewebe

μ_G(E)

linearer Schwächungskoeffizient Drüsengewebe/cm^–1

μ_F(E)

linearer Schwächungskoeffizient Fettgewebe/cm^–1

α = xG/H = bG/(ρGH) (#2a) 1 – α = xF/H = bF/(ρFH) (#2b) β(E) = μF(E)/μG(E) (#2c) Second, it is assumed that, given the resulting spectral distribution, N _U (E) and given Mamma layer thickness H, which is determined by the distance of the compression plates 5 the weakening of the detector signal (of primary X-ray radiation, without stray radiation) as a function of the tissue portion of glandular tissue (fat tissue) predicted (possibly validated by measurements) is present, that is, the following function is in tabular form where:

With

H

Layer thickness of the mom 6

x _G

Layer thickness glandular tissue / cm

x _F = H - x _G

Layer thickness fatty tissue / cm

ρ _G , ρ _F

Dense glandular or fatty tissue [g / cm ³ ]

b _G = ρ _G x _G

Mass allocation of glandular tissue [g / cm ² ]

b _F = ρ _F x _F

Mass allocation of fatty tissue

μ _G (E)

linear attenuation coefficient gland tissue / cm ^-1

μ _F (E)

linear attenuation coefficient adipose tissue / cm ^-1

α = x G / H = b G / (Ρ G H) (# 2a) 1 - α = x F / H = b F / (Ρ F H) (# 2b) β (E) = μ F (E) / μ G (E) (# 2c)

It is assumed that the compressed mom 6 the layer thickness H between the compression plates 5 fully filled. This condition is according to 4 in the range of a few centimeters near a breast tip 26 and outside in the field of un-weakened radiation is no longer fulfilled. These image field areas must be treated separately as part of a precorrection, as will be explained in detail below, for example by suitable extrapolation of the tissue layer thickness H to 0.

For computational reasons, the logarithmic attenuation signal is more useful than the non-logarithmic attenuation function F in equation (# 2):

tabelliert zur Verfügung steht.The function f _H is monotonic and continuous and thus invertible, for example by inverse interpolation. Therefore, it can be assumed that also the inverse function

tabulated is available.

Third, it is assumed that the so-called mamma SBSF atlas 19 is present, because the method described here is based on the knowledge of the respective SBSFs 20 (= Scatter beam spread functions), which are also referred to as scatter beam propagation functions. An SBSF 20 in each case describes the spatial intensity distribution of the scattered radiation on the X-ray detector designed as an area detector 8th for a thin X-ray (beam) of the X-radiation corresponding to the scattering object (mamma) 1 permeates at a given location. The SBSF 20 depends on recording parameters and object parameters.

Recording parameters are, for example, the tube voltage, which influences the photon emission spectrum, which also depends on the anode material, the pre-filtering, the air gap, the so-called SID (= source-image distance), the Kollimie tion (detector insertion), the spectral response of the X-ray detector 8th and the presence or absence of a scattered radiation grid.

Object parameter is on the one hand the layer thickness H of the mamma 6 and on the other hand the different proportion of fatty and glandular tissue along the propagation direction of the X-radiation 3 ,

It is assumed that the SBSFs 20 for the most important recording and object parameters available, that is, a pre-prepared spreadsheet, called the mamma SBSF atlas 19 , is present, with the help of which it is possible for the specific given receiving conditions for each proportion of fat and glandular tissue (scattering material distribution) along an X-ray beam, the associated SBSF 20 be determined with sufficient precision, for example by interpolation in the mammary SBSF atlas 19 or by semi-empirical conversions of parameters of which the SBSF is only weakly dependent or for which functional dependencies are known, such as the SID.

The Mom SBSF Atlas 19 is prepared in advance by means of Monte Carlo simulation calculations. The Monte Carlo simulation allows the physical processes of absorption and multiple scattering (predominantly coherent scattering in the low energy range of mammography) to pass through the scattering object, in particular the mamma 6 to adequately model, taking into account the conditions of reception (anode material, filter, voltage, air gap, SID, field size, possibly anti-scatter grid). This is the decisive advantage of the Monte Carlo method compared to analytical simulation models, which are usually limited to single scattering and usually still simplifications and approximations are introduced to reduce the effort. The calculation of scattering distributions on the basis of a Monte Carlo Si simulation is known in the art and as such is not the subject of the application.

Description of the individual Steps:

• 0. Leerbild-Kalibrierung und Bestimmung des effektiven Schwächungssignals (wobei schon eine einfache pauschale Streustrahlungs-Vorkorrektur empfehlenswert ist);
• 1. Bestimmung des Anteils von Drüsengewebe und Fettgewebe;
• 2. Schätzung der Streustrahlungsverteilung (genaueres SBSF-Modell);
• 3. Schätzung der Primärstrahlungsverteilung (Streustrahlungs-Korrektur);
• 4. iterative Wiederholung ab Schritt 1. oder Ende.

The scattered radiation correction is subdivided into the following individual process steps, which can be repeated in an iterative cycle:

• 0. Blank image calibration and determination of the effective attenuation signal (even a simple blanket scatter pre-correction is recommended);
• 1. Determination of the proportion of glandular tissue and fatty tissue;
• 2. Estimation of the scattered radiation distribution (more exact SBSF model);
• 3. estimate of the primary radiation distribution (scattered radiation correction);
• 4. iterative repetition from step 1 or end.

The steps 0 and 1 are to be carried out for each measuring beam, that is to say for each pixel (j, k), hereinafter the term pixel for both the detector pixels 9 as well as for multiple detector pixel detector areas is used.

Process step 0: I ₀ calibration and attenuation signal with precorrection

Let I ₀ (j, k) be the blank image, which is equal to the measured intensity distribution in the beam path without scattering object, I (j, k) the measured intensity distribution with scattering object (Mamma), then the effective attenuation signal for total radiation, that is, the Superposition of primary and secondary (= scattered) radiation, given by: T (j, k) = I (j, k) / I 0 (j, k). (# 5a).

In general, with regard to step 1, it will be expedient to carry out a precorrection of the scattered radiation background, which is to be designated S ⁽⁰⁾ , already here. Methods for estimating S ⁽⁰⁾ are added below. S ⁽⁰⁾ can be location-dependent, but is constant in the simplest case. The pre-correction already provides an estimate of the primary attenuation signal (normalized primary intensity) P (0) (j, k) = T (j, k) - S (0) (# 5b).

Process Step 1: Estimation of Specific tissue shares

und die Massenbelegung Drüsengewebe [g/cm²]: bG = αρGH (#6a)sowie die Massenbelegung Fettgewebe: bF = (1 – α)ρFH (#6b) Assuming at first that P (j, k) represents only primary radiation without stray radiation, equation (# 4) and (# 3) give for the proportion of glandular tissue:

and the mass population of glandular tissue [g / cm ² ]: b G = αρ G H (# 6a) as well as the mass allocation of fatty tissue: b F = (1 - α) ρ F H (# 6b)

There The above assumption, strictly speaking, is not true, is one iterative approach required. This is related with the versions even closer to step 4 be executed.

Process step 2: possible correct estimate the scattered radiation distribution over the whole projection picture

To belong to this process step several sub-steps:

2.1 Lookup in the Mum SBSF Atlas

The generation of the SBSF atlas 19 will be described in detail below.

For each beam associated with a pixel (j, k), α (j, k) was calculated in method step 1. To the calculated value of α (j, k) and H and other parameters such as air gap (air gap), spectrum and other parameters then the associated SBSF 20 from the mamma SBSF atlas 19 generally determined by interpolation:
SBSF ((λ _x , λ _y ); α; H; air gap, voltage, filter, detector, ....)

SBSF is a two-dimensional function, or rather a two-dimensional array (data array), depending on the row and column coordinates on the X-ray detector 8th , Every SBSF 20 is focused on a center, namely the respective beam or rather on the relevant pixel with the coordinates (0,0) and falls off at a distance from the beam center strongly. The distance from the center in both coordinate directions is indicated by an index pair (λ _x , λ _y ). The SBSF 20 is a kind of point or line image function, where the point or line is actually the ray.

To mark the interpolation, we use the notation: SBSF I ((Λ x , λ y ); α) with α = α (j, k) (# 7a)

This SBSF 20 is attached to the pixel (j, k) with its center (λ _x , λ _y ) = (0, 0) to a certain extent. Thus, for each ray or pixel (j, k) we get the SBSF with which that ray or pixel contributes to the total scattering intensity distribution across the detector surface; this contribution we denote by ΔS: .DELTA.S (j, k) (λ x , λ y ) = SBSF I ((Λ x , λ y ); α (j, k)) (# 7).

2.2 Integration of the scattered radiation distribution over the detector

The Contributions ΔS must now over all Pixels are integrated.

The SBSFs 20 are normalized to the attenuation = 1 of the relevant ray (pixel). When summing all contributions must therefore be multiplied by the actual attenuation.

We hold a pixel (j, k) and consider all the pixels (j ', k') in terms of their contribution to the total scattered radiation in (j, k). The SBSF attached to the pixel (j ', k'), as it were, then carries the following equation (# 7): .DELTA.S (j, k) (λ x , λ y ) · P (j ', k') with λ x = j - j ', λ y = k - k '(# 8) at the point (j, k) at.

With (# 7) to (# 8) one for one the scattered radiation at the location (j, k):

This applies to arbitrary pixels (j, k) and thus by equation (# 9) the whole Stray radiation distribution described.

2.3 low-pass filtering

The Stray radiation distribution is multiple because of the generating them Scattering processes in the body relatively smooth and therefore has a low frequency Fourier spectrum on. To possibly through the previous processing steps to eliminate induced high-frequency error components is one 2-dimensional smoothing to recommend.

Process Step 3: Scattering Correction

In fact, the data available are initially uncorrected, that is, measurement-based data representing the superposition of primary radiation 10 (direct, unscattered radiation) and secondary radiation 11 (= Include scattered radiation.

T

gemessene (normierte) Verteilung der totalen Strahlung

P

zunächst unbekannte, aber gesuchte (normierte) Primärstrahlung 10

S

unbekannte, aber mit dem vorgeschlagenen Modell geschätzte (normierte) Sekundärstrahlung 11.

After normalization according to equation (# 5a): T = P + S, (# 10) with the meanings:

T

Measured (normalized) distribution of total radiation

P

initially unknown, but sought (normalized) primary radiation 10

S

unknown (normalized) secondary radiation estimated but with the proposed model 11 ,

By normalization is meant the division by the intensity distribution I ₀ (j, k) without scattering object.

Equation (# 9) directly yields a subtractive stray radiation correction: P (j, k) = T (j, k) - S (j, k) (# 11) to estimate the primary radiation distribution.

Another correction, resulting in cases of a relatively large amount of secondary radiation 11 recommends is the multiplicative stray radiation correction: P = T / (1 + S / P) (# 12)

you notice that the corrections in equation (# 11) and equation (# 12) are only approximate and do not give identical results. For S / T << 1, however, (# 11) changes to (# 12).

Process step 4: iteration

In Equation (# 11) and (# 12), on the right side, the scattered radiation term S occurs, which in turn is calculated by Equation (# 9); Equation (# 9), however, is defined by means of the (unknown) primary radiation P, which in turn appears on the left side of Equation (# 11) and (# 12) and should only be computed by one of these equations. Thus, P appears on both the left and right sides of Equation (# 11) and (# 12). Such implicit equations have to be solved iteratively. We write for S in equation (# 9): S = S (P) (# 13a)

Equation (# 11) then reads: P = T - S (P) (# 13b)

The iteration is done for the subtractive method as follows:
Iteration start with precorrection, which will be described in more detail below: P (0) = T - S (0) (# 5b) = (# 14a)

iteration: P (n + 1) = T - S (P (N) ), n + 1>0;(# 14b)

For the multiplicative method, the iteration is performed as follows:
Iteration start with precorrection, which will be described in more detail below: P (0) = T - S (0) (# 5b) = (# 15a)

iteration: P (n + 1) = P (N) T / (P (N) + S (P (N) )), n + 1> 0. (# 15b)

The Consequence of the iterations is canceled each time, if the result between step n and n + 1 changes only little. In many cases already enough one cycle (n = 1).

SNR improvement through statistical estimation: ML and Bayes methods

Interestingly, the multiplicative correction method (# 15b) can be derived from a statistical estimation approach using the maximum likelihood principle (ML). In the relevant literature, a simple convolution model is used for the scattering operator, S (P) in equation (# 13a), for example in AH BAYDUSH, CE FLOYD: Improved image quality in digital mammography with image processing. In: Med. Phys., Vol. 27, July 2000, pages 1503 to 1508. However, the ML principle can in principle be applied independently of the specific scattering model, in particular also in the case of the scattering model described here.

One Method based on the ML principle has the property that usually the SNR (= signal to noise ratio) improved after a few iterations but, if the iterations continue, the noise will be uncontrolled increases and the SNR worsens again. To run away from this to counteract the ML algorithm, Bayesian estimation methods are recommended. resulting in algorithms that differ from equation (# 15b) through a stabilizing additional Term on the right side differ. The effect of the additional term on convergence speed, SNR and the tradeoff between Noise and spatial resolution can be controlled by parameters.

pre-corrections

In the previous comments on method steps 1 and 2.1, there equation (# 6) and (# 7), it was assumed that the compressed mom 6 the layer thickness H between the compression plates 5 Fills completely and that the function f _H ^-1 can be evaluated. This condition is according to 4 in the range of a few centimeters near a breast tip 26 and outside in the area of unattenuated X-radiation 3 not fulfilled anymore. These image field areas must be treated separately as part of a precorrection. In the area of unattenuated X-radiation 3 outside the mom 6 the effective attenuation signal according to equation (# 5a) must theoretically = 1, but will generally be> 1 due to the presence of stray radiation. The difference ΔT (j, k) = I (j, k) / I 0 (j, k) - 1 (if> 0) must therefore be considered a scattered radiation pre-correction S (0) = ΔT in the image area outside the mom 6 subtracted from.

From the normal image area of fully compressed mom 6 to the area near the breast tip 26 appropriate extrapolation of the tissue layer thickness of H to 0 should be carried out. In this image area, therefore, equations (# 2), (# 6), and (# 7) are generally considered to be H variable.

Optionally, a segmentation in 3 image areas according to K. NYKÄNEN, S. SILTANEN: X-ray scattering in full field digital mammography. In Med. Phys., Vol. 30 (7), July 2003, pages 1864 to 1873 are made.

In the normal image area with constant tissue layer thickness H, a scattered radiation pre-correction can look like this: Since there is no evaluation of the tissue components (glandular / fatty tissue), one can first assume 100% fat. Because of the lower density of fat (0.92 vs. 0.97 g / cm ³ in glandular tissue), the scattered radiation is underestimated, but for a zero-order correction, this estimate is much better than no correction at all. In Equation (# 7) and the following equations, α = 0 is used, and thus the scattered radiation kernel SBSF becomes location independent, particularly independent of the pixel index (j, k), and Equation (# 9) reduces to a true convolution.

The equations (# 7- # 9) are simplified as follows: We omit the index at ΔS (j, k) and write ΔS ⁽⁰⁾ : .DELTA.S (0) (λ x , λ y ) = SBSF I ((Λ x , λ y ,); α = 0) (# 16a); instead of P, set T (# 9) according to equation (# 5a) in (# 9):

there means ** a 2-dimensional convolution.

The precorrection then supplies according to equation (# 5b): P (0) = T - S (0) = T - (ΔS (0) ** T) (# 16c)

Creating the mamma SBSF atlas

In the concept of SBSF, one is interested in the distribution of the scattered radiation generated in the scatterer in the detector plane, if in accordance with 4 the (unscattered) primary beam (that is, a mini beam cone 27 ) exactly on a detector pixel 9 is focused. Do this one at a time for each detector pixel 9 and sums all associated SBSFs 20 on, then you get the entire scattered radiation distribution in the event that the entire detector surface is illuminated - and not just individual detector pixels 9 ,

The Mom SBSF Atlas 19 the scatter beam propagation functions 20 (= Scatter-Beam-Spread-Functions (= SBSF)), as already mentioned above in connection with the third prerequisite and the method step 2 described (on the intensity of the primary radiation 10 in the detector pixel 9 normalized) scattered radiation intensity distributions (assuming focussing of the mini beam cone 27 to exactly one detector pixel 9 ) depending on a variety of different parameter configurations: SBSF ((λ x , λ y ); α; H; Air gap, voltage, filter, detector, ...) (# 17) Also contains the dependence of the X-ray energy spectrum on the tube voltage, the pre-filtering, the radiation-sensitive detector material, for example, the type of scintillation crystal, and the dependence on the presence or absence of a scattered radiation grid and optionally the dependence on the type of antiscatter grid and the dependence of others parameters.

Below is the now Creating an SBSF series explains:

First, those for the underlying mammography device 1 defining characteristic parameters: SID, air gap, X-ray tube anode material (and associated emission spectra), detector material, pre-filter materials (for example, compression plates), and other parameters. Then comes the compression thickness H, the voltage, the spectral filters used and other variables, generally to optimize the image quality, the voltage and optionally the spectral filters (thick) depending on the compression thickness H are modified.

For this parameter configuration then the parameter α, which is the tissue composition according to equation (# 2a), between 0 (only fat) and 1 (only gland tissue) varies: The calculation with the proven Monte Carlo method results in a set of different SBSFs 20 where each α value is an SBSF 20 is assigned.

Then the tissue thickness H is varied between> 0 and up to about 10 cm, and for each H again another set of SBSFs 20 calculated. Furthermore, the voltage and the spectral filters can be varied, with the variation being coupled to H or else independent of H. In the latter case, however, there is a multiple of variation possibilities. Incidentally, the calculation can be continued for all parameter combinations.

• • Vernachlässigung der Divergenz der Strahlen der Röntgenstrahlung 3 auf Grund der Kegelstrahl-Geometrie, indem näherungsweise Parallelstrahlgeometrie angenommen wird; das ist dadurch gerechtfertigt, dass in der Regel SID >> H ist; dadurch erreicht man, dass die SBSF 20 bei gleicher Konfiguration des Strahls orts- und pixelunabhängig bleibt; unter gleicher Konfiguration soll verstanden werden, dass für jedes Pixel die Materialverteilung längs des Ministrahlkegels 27 und in der seitlichen Nachbarschaft gleich ist.
• • Zur Verbesserung der Statistik beim Monte-Carlo-Verfahren und zur Verringerung des Rechenaufwands werden für die Berechnung der SBSFs 20 um etwa eine Größenordnung größere Pixel (z.B. 1 × 1 mm² oder 2 × 2 mm²) verwendet als die tatsächlichen Detektorpixel 9 (≤ 0.1 mm); dies ist zu rechtfertigen durch das niederfrequente Fourier-Spektrum der räumlichen Streustrahlungsverteilung.
• • Die Aufeinanderfolge von Fett- und Drüsengewebe wird ersetzt durch ein Gemisch; zwar hängt die Streustrahlung (bei gleicher gesamter Massenbelegung und Weglänge) davon ab, ob sich das dichtere Gewebe näher beim Röntgendetektor 8 oder näher bei der Strahlungsquelle 2 befindet; gemäß J.M. DINTEN und J.M. Volle: Physical model based restoration of mammographies. In Proc. SPIE, Vol. 3336, 1998, 641–650 können aber die unter mammographischen Bedingungen auftretenden Unterschiede vernachlässigt werden.

For the calculation of SBSFs 20 Simplifications can be made that are well justified:

• • Neglecting the divergence of the X-ray beams 3 due to the cone-beam geometry, assuming approximately parallel-beam geometry; this is justified by the fact that usually SID >>H; This achieves that the SBSF 20 remains spatially and pixel-independent with the same configuration of the beam; Under the same configuration is to be understood that for each pixel, the material distribution along the mini beam cone 27 and equal in the lateral neighborhood.
• To improve Monte Carlo statistics and reduce computational effort, the calculation of SBSFs 20 uses pixels larger by about one order of magnitude (eg, 1 × 1 mm ² or 2 × 2 mm ² ) than the actual detector pixels 9 (≤ 0.1 mm); this is justified by the low-frequency Fourier spectrum of the spatial scattered radiation distribution.
• • The sequence of fatty and glandular tissue is replaced by a mixture; Although the scattered radiation (with the same total mass occupancy and path length) depends on whether the denser tissue is closer to the X-ray detector 8th or closer to the radiation source 2 is located; according to JM DINTEN and JM Full: Physical model based restoration of mammographies. In Proc. SPIE, Vol. 3336, 1998, 641-650, however, can neglect the differences that occur under mammographic conditions.

advantages

The solution proposed here has the following advantages:
The method may optionally be integrated into existing mammography devices without mechanical modification.

Further it is a procedure that, on the one hand, is adequacy the physical modeling with the Monte Carlo method divides, on the other hand - because all elaborate bills as far as possible in the Carried out in advance and the necessary data are stored in tables - ultimately with relatively little computational effort for the scattered radiation correction gets along.

The Model accuracy of the scattered radiation correction described here is in principle larger than that of the known (analytical) physical models, since dispensed with a series of simplifying assumptions and approximations can be.

The options the here proposed stray radiation correction go over the Possibilities of long known convolution / deconvolution method out. If apart from the specific technical embodiment of the method and the method is considered from the mathematical point of view Thus, in the mathematical sense, the method can be considered as a generalization of the considered a long-known convolution / deconvolution method become. Therefore lets on the one hand, by approximations and no accuracy, convert into this type class and then shares their advantages, for example the possibility of using the so-called FFT (= fast Fourier transformation). on the other hand However, the method described here can be improved in terms of SNR improvement also be extended, for example, by the iterative multiplicative Algorithm is extended in the direction of statistical Bayes estimate.

In this context, it should again be noted that only the prediction of the SBSFs 20 allows the implementation of the method described herein in full generality.

embodiments

Embodiment 1

In this embodiment, the scattering correction, as described above in connection with equations (# 5) - (# 9) and (# 13) - (# 15), is done with homogeneous location-dependent scatter beam propagation functions 20 (= SBSF). When creating the scatter beam propagation functions 20 For the sake of simplification, it is assumed that the tissue distribution characterized by the proportion α (j, k) of glandular tissue along the beam leading from the source to the detector pixel corresponds correspondingly 4 at right angles to the beam, ie parallel to the compression plates 5 , continues unchanged homogeneous. Thus, with respect to the scattering contribution of the beam in the pixel (j, k), it is assumed that the tissue composition in the lateral vicinity of the beam does not change abruptly. Although this is no longer true at the Mammarand, but there could be a special treatment.

It should be noted, however, that the actual location-dependent inhomogeneity of the tissue composition is taken into account by a specific glandular tissue component α (j ', k') specific to each pixel (j ', k') and a specific scattering contribution dependent thereon. The SBSFs 20 are therefore usually different for each pixel.

Embodiment 1a

at this embodiment 1, the method becomes substantially the same as Embodiment 1 executed.

However, some simplifications are made:
For each predefined layer thickness and the other parameters, such as voltage and prefiltering, a common SBSF 20 used for all pixels. In this case, the SBSF 20 thus chosen location independent. The selection can be made for example by a suitable averaging over the occurring tissue compositions. ΔS in Equation (# 7) and (# 9) then becomes independent of the pixel index (j, k); the double index (j, k) can be omitted - similar to the equations (# 16a) to (# 16c).

Important is that the integral in equation (# 9) turns into a true convolution, which are efficiently executed by FFT (= fast Fourier transform) can.

Embodiment 1b:

at this embodiment 1, the method becomes substantially the same as in the embodiment 1 executed.

In However, in this case, a uniform convolution kernel (for all layer thicknesses) for the Stray radiation calculation used. That with a small layer thickness relatively less scattered radiation arises than with a large layer thickness, must by scaling factors, by the layer thickness and others Parameters, such as voltage and filtering, are taken into account become.

In comparison to the exemplary embodiment 1a, approximately the same amount of computation is required for the exemplary embodiment 1b. For this, in this embodiment, much less storage space for storing the mamma SBSF atlas 19 necessary.

Comments on the embodiments 1a and 1b:

In general, the simplified embodiments 1a and 1b share the property that the convolution models for the scattered radiation can be inverted by means of the Fourier transform. Then one speaks of deconvolution. Of the conventional deconvolution methods, the embodiments described here differ by the use of one or more scatter beam spread functions previously obtained with the aid of a Monte Carlo simulation 20 ,

For the execution of a deconvolution, see a publication by JA SEIBERT and JM BOONE: X-ray scatter removal by deconvolution. In Med. Phys., Vol. 15, 1988, pages 567-575. Reference is also made to the recent publication P. ABBOTT et al .: Image deconvolution as an aid to mammographic artifact identification I: basic techniques. In: Proc. SPIE, Vol. 3661, 1999, pages 698 to 709, which deals with a deconvolution with regularization techniques for noise suppression. Another deconvolution method with thickness-dependent convolution is described in DG TROTTER: Thickness-Dependent Scatter-Correction Algorithm for Digital Mammography. In: Proc. SPIE, Vol. 4682, 2002, pages 469-478. In this method, an iteration with relaxation is performed.

Embodiment 2

In this embodiment, the process is performed substantially as in Embodiment 1, but with scattered beam propagation functions 20 worked, which have been calculated for an inhomogeneous medium.

In 5 For example, the case shown is a mammary area 28 has a different composition than a surrounding mammal area 29 ,

This can take into account that the SBSF 20 not just the tissue composition along the mini-beam cone focused on the detector pixel 27 but also of the tissue composition in the lateral neighborhood into which photons can be scattered and redistributed in the direction of the pixel. However, the range of action of the lateral neighborhood is not very large because of the mean free path <~ 2 cm of photons in the mammographic energy range between about 20 and 40 keV. It would therefore be sufficient to assume the tissue composition as homogeneous in a lateral hemisphere, but generally different from the mini-beam cone 27 , The consideration of inhomogeneous SBSFs 20 with differences between jet and neighborhood should play a role especially at the Mammarand.

This embodiment therefore represents a generalization of the embodiments 1, 1a and 1b described above, since in this case the SBSFs 20 depend not only on a tissue parameter α, but also on a newly introduced environmental tissue parameter γ. In this case, the mamma SBSF atlas 19 thus have an additional dimension.

For the sake of clarity, the different characteristics of the exemplary embodiments 1, 1a, 1b and 2 are compared in the following table:

Embodiment 3

The method described here also on the expert known as the so-called dual-energy method apply. In the so-called dual-energy process, this is especially true used in mammography or in bone densitometry become temporally parallel with two different energy spectra Recordings made. The pictures with different energy spectra be through two different voltages and as possible also accomplished various spectral filtering, so that the two measurements effectively corresponding spectral ranges as possible little overlap. By a computing process, which is essentially based on the solution in general non-linear system of two the two spectra based equations, then can be compared to a recording with an energy spectrum finer tissue differentiation be achieved. For the arithmetic process to succeed, the Stray radiation as possible be eliminated, otherwise due to the scattered radiation components may induce artifacts stronger are considered the actual tissue image.

Because of The difference of scattered radiation in both spectra is one powerful Stray radiation correction therefore for the quality the dual-energy method is crucial.

The proposed stray radiation correction method is also in this Context applicable. The geometric parameters are for both Recordings same, but the spectral-dependent parameters are different.

The correction is to be carried out for each of the two images according to the scheme described, with the only difference being that different SBSFs correspond to the different spectra 20 must be used.

Claims (19)

Device for projection radiography, in particular for mammography, with radiation ( 3 ) emitting radiation source ( 2 ), a detector ( 8th ) and a detector ( 8th ) downstream evaluation unit ( 12 ) based on the detector ( 8th ) ( 17 ) a scattering material distribution of an object to be examined ( 6 ) and determined in dependence on the distribution of scattering material from a data memory ( 19 ) Scatter information ( 20 ) and based on the scatter information ( 20 ) the projection data ( 17 ) with regard to the scattered radiation component ( 11 ), characterized in that the scatter information ( 20 ) are determined by Monte Carlo simulations, which calculate the interactions of the photons, each with a scattering material distribution, for different scattering material distributions. Apparatus according to claim 1, characterized in that the scattering information scattering distributions ( 20 ), which are a scattering caused by the distribution of the radiation source ( 2 ) and directed to a specific image area radiation ( 14 ) on adjacent image areas. Device according to claim 2, characterized in that the scattering distributions ( 20 ) with the intensity of the detector ( 8th ) incident unscattered primary radiation ( 10 ) are scalable. Device according to one of claims 1 to 3, characterized in that the evaluation unit ( 12 ) for different image areas of a projection image ( 17 ) for the respective scattering material distribution specific scattering information ( 20 ) evaluates. Device according to one of claims 2 to 4, characterized in that the evaluation unit ( 12 ) a scattered radiation distribution ( 11 ) in an image area of the projection image ( 17 ) determined by the evaluation unit ( 12 ) for each image area the scattered radiation contributions ( 11 ) of the surrounding image areas is calculated and added. Device according to one of claims 2 to 4, characterized in that the evaluation unit ( 12 ) a scattered radiation distribution ( 11 ) in an image area of the projection image ( 17 ) determined by the evaluation unit ( 12 ) the primary radiation distribution with a scattering distribution ( 20 ) folds. Apparatus according to claim 5 or 6, characterized in that the evaluation unit ( 12 ) the unscattered primary radiation ( 10 ) is determined by solving the implicit equation P + S (P) = T, where P is the distribution of the unscattered primary radiation ( 10 ), S (P) that of the unscattered primary radiation ( 10 ) dependent secondary radiation distribution ( 11 ) and T is the measured total radiation distribution in the projection images ( 17 ). Device according to one of claims 1 to 8, characterized dadudurch that the evaluation unit ( 12 ) for a first approximate determination of the scattering material distribution of the object to be examined ( 6 ) in the projection image ( 17 ) contained in scattered radiation ( 11 ) on the basis of scatter information associated with a typical scatter material distribution. Device according to one of claims 1 to 8, characterized in that the evaluation unit ( 12 ) iteratively executes the processing steps of any one of claims 1 to 8. Device according to one of claims 1 to 9, characterized in that the scatter information ( 20 ) are calculated on the assumption of a homogeneous scattering material distribution in the beam direction. Device according to one of claims 1 to 10, characterized in that in the data memory ( 19 ) Scatter information ( 20 ) are stored, which the outer contour ( 26 ) of the object to be examined ( 6 ) consider. Device according to one of claims 1 to 11, characterized in that in the data memory ( 19 ) Scatter information ( 20 ) are stored, which are calculated on the assumption of a homogeneous distribution of scattered material transverse to the beam direction. Device according to one of claims 1 to 12, characterized in that in the data memory ( 19 ) Scatter information ( 20 ) are stored, which are determined taking into account a non-inhomogeneous distribution of scatter material transverse to the beam direction. Device according to one of claims 1 to 13, characterized in that in the data memory ( 19 ) Scatter information as a function of parameters of the radiation source ( 2 ) are stored. Device according to one of claims 1 to 14, characterized in that the evaluation unit ( 12 ) the scattered radiation components ( 12 ) at selected nodes and the correction values for individual detector elements ( 9 ) determined by interpolation between the nodes. Device according to one of claims 1 to 15, characterized in that the object to be examined ( 6 ) in a compression device ( 5 ) is compressible and that the evaluation unit ( 12 ) the spatial design of the object to be examined ( 6 ) facing surfaces of the compression device ( 5 ) for determining the path length of the radiation ( 3 ) through the object to be examined. Method for scattered radiation correction in projection radiography, in which by means of a detector ( 8th ) a scattering material distribution of an object to be examined ( 6 ) from an evaluation unit ( 12 ) is approximately determined, and in which, depending on the scattering material distribution scattering information ( 20 ) from the evaluation unit ( 12 ) from a data store ( 19 ), on the basis of which the projection data ( 17 ) with regard to the scattered radiation component ( 12 ), characterized in that by a Monte Carlo simulation determined scatter information ( 20 ) can be used by the multiple interactions between the photons and the object to be examined ( 6 ). Method for obtaining scatter information for scattered radiation correction, in which, in a Monte Carlo simulation, the path of a multiplicity of photons through an object to be examined ( 6 ), characterized in that a plurality of scattering distributions ( 20 ) for various parameters of the scattering material distribution of the object to be examined ( 6 ) and for various parameters one for the examination of the object ( 6 ) used device ( 1 ) and stored in tabular form. Method according to claim 18, characterized in that the scattering distributions ( 20 ) for different geometries of the object to be examined ( 6 ) are calculated and stored.